Image Super-Resolution Reconstruction System and Method

1. An image super-resolution reconstruction method, comprising: an edge detection step of performing an edge detection on low-resolution images to be processed to obtain edge pixel frames; a frame amplification step of amplifying the edge pixel frames so that each amplified image is the double of the original edge pixel frame in size in both horizontal direction and vertical direction, without changing the detected edge pixel information; and a pixel compensation step of compensating for the interpolated interpolation pixels according to different pixel edges to obtain a high-resolution images; wherein, the pixel compensation step further comprises: searching, based on the central interpolation pixel in five interpolation pixels to be compensated which is located at the center of a pixel block, the information of the edges pixels contained in the four pixels adjacent to the central interpolation pixel in set four directions, and then compensating for the five interpolation pixels accordirato corresponding rules, and wherein, the rules are related to the number of the edge pixels contained in the four pixels,the edge directions, and/or brightness of the pixels: wherein, when the set directions are selected in a range from 45° to 90° , if the assigned value to one of the five interpolation pixels currently compensated is also from interpolation pixels previously compensated, and the pixel block compensated previously is located above the interpolation pixel, then the assigned value of the interpolation pixel is the value currently compensated, and, if the assigned value to one of the five interpolation pixels currently compensated is also from one interpolation pixels previously compensated, and the pixel block compensated previously is located at the left side of the interpolation pixel, then the assigned value of the interpolation pixel is the value previously compensated; wherein, when the set directions are selected in a range from 90° to 180°, if the assigned value to one of the five interpolation pixels currently compensated is also from one of interpolation pixels previously compensated, and the pixel block compensated previously is located above the interpolation pixel, then the assigned value of the interpolation pixel is the value previously compensated, and if the assigned value to one of the five interpolation pixels currently compensated is also from one of interpolation pixels previously compensated, and the pixel block compensated previously is located at the left side of the interpolation pixel, then the assigned value of the interpolation pixel is the value currently compensated; and wherein, when the set directions are selected from outside of a range of 45° to 180°, if the assigned value to one of the five interpolation pixels currently compensated is from one of interpolation pixels previously compensated, and the pixel block compensated previously is located above or at the left side of the interpolation pixel, then the assigned value of the interpolation pixel is the value previously compensated.

2. The method according to claim 1 , wherein, in the edge detection step, the edges of each low-resolution image to be processed are further detected from four different directions by adopting a Sobel operator.

3. The method according to claim 1 , wherein, the rules comprise: if all the four pixels are edge pixels, the edge are in diagonal directions, and the difference of brightness between the two pixels in the edge directions and the other pixel is less than a set threshold value, then the interpolation compensation is performed by virtue of the following steps: assigning the average of brightness values of the three pixels to the central interpolation pixel located at the center of the pixel block; assigning the brightness value of other pixel except for the three pixels to the two interpolation pixels adjacent to the other pixel; and assigning the average of brightness values of each of the two pixels in the edge directions and the other pixel to the interpolation pixel therebetween.

4. The method according to claim 1 , wherein, the rules comprise: if all the four pixels are edge pixels, the difference of brightness values between the two pixels in each row is less than a set threshold value, and the difference of the averages of brightness values of the two adjacent rows is greater than the set threshold value, then the interpolation compensation is performed by virtue of the following steps: assigning the brightness average of two pixels in each row to the interpolation pixels in the corresponding row; assigning the brightness average of two pixels in the lower row to the central interpolation pixel located at the center of the pixel block; and assigning the brightness value of each pixel in the two pixels in the lower row to the interpolation pixels in the column where each pixel in the two pixels is located.

5. The method according to claim 1 , wherein, the rules comprise: if all the four pixels are edge pixels, and the brightness difference value between the two pixels in each column is less than a set threshold value and the difference of the brightness averages of the left column and the right column is greater than the set threshold value, then the interpolation compensation is performed by virtue of the following steps: assigning the brightness average of two pixels in each column to the interpolation pixels in the column; assigning the brightness average of two pixels in the left column to the central interpolation pixel located at the center of the pixel block; assigning the brightness value of each pixel in the two pixels in the left column to the interpolation pixels in the row where the pixel is located.

6. The method according to claim 1 , wherein, the rules comprise: if all the four pixels are edge pixels, and the difference between the brightness values of the four pixels is less than a set threshold value or greater than the set threshold value, then the interpolation compensation is performed by virtue of the following steps: assigning the average of the brightness values of two pixels in each column to the interpolation pixels in the column; assigning the average of the brightness values of two pixels in each row to the interpolation pixels in the row; and assigning the average of the brightness values of the four pixels to the central interpolation pixel located at the center of the pixel block.

7. The method according to claim 1 , wherein, the rules comprise: if the number of edge pixels in the four pixels is three, and the edges are in a diagonal direction, then the interpolation compensation is performed by virtue of the following steps: assigning the average of the brightness values of the three pixels to the central interpolation pixel located at the center of the pixel block; assigning the brightness values of other non-edge pixels except the three pixels to the two interpolation pixels adjacent to the non-edge pixel; and assigning the average of the brightness values of each of the two pixels in the edge directions and another edge pixel except for the two edge pixels to the interpolation pixels therebetween.

8. The method according to claim 1 , wherein, the rules comprise: if the number of edge pixels in the four pixels is two, and the edges are in a diagonal direction, then the interpolation compensation is performed by virtue of the following steps: assigning the average of the brightness values of the two edge pixels to the central interpolation pixel located at the center of the pixel block; and assigning the brightness value of each of the other two non-edge pixels to the interpolation pixels adjacent to the non-edge pixel.

9. The method according to claim 1 , wherein, the rules comprise: if the number of edge pixels in the four pixels is two, and the edges are in directions of columns, then the interpolation compensation is performed by virtue of the following steps: assigning the brightness average of the two edge pixels to the interpolation pixels in the column where the two pixels are located; assigning the average of brightness values of other two non-edge pixels to the central interpolation pixel located at the center of the pixel block; and assigning the brightness value of each of the other two non-edge pixels to the interpolation pixels adjacent to the non-edge pixel.

10. The method according to claim 1 , wherein, further including a step of compensating for the remaining interpolation pixels by adopting a bicubic interpolation method.

11. An image super-resolution reconstruction system, comprising: an edge detection means, for performing an edge detection on low-resolution images to be processed to obtain edge pixel frames; a frame amplification means for amplifying the edge pixel frames so that each amplified image is the double of the original edge pixel frame in size in both horizontal direction and vertical direction, without changing the detected edge pixel information; and a pixel compensation means for: compensating for the interpolated interpolation pixels according to different pixel edges to obtain a high-resolution image; searching, based on the central interpolation pixel five interpolation pixels to be compensated which is located at the center of a pixel block, the information of the edge pixels contained in the four pixels adjacent to the central interpolation pixel in set four directions, and then compensating for the five interpolation pixels according to corresponding rules, wherein the rules are related to the number of the edge pixels contained in the four pixels, the edges directions, and/or brightness of the pixels, and recording the five interplation pixels compensated every time, selects sets directions so as to conduct a mutual exclusivity analysis, for judging whether the assigned values to the five interpolation pixels currently compensated are mutually exclusive with the assigned values previously compensated, and processing the interpolation pixels which are mutually exclusive; wherein, when the set directions are selected in a range from 45° to 90°, if the assigned value to one of the five interpolation pixels currently cormpensated is also from interpolation pixels previously compensated, and the pixel block compensated previously is loated above the interpolation pixel, then the assigned value of the interpolation pixel is the value currently compensated, and, if the assigned value to one of the five interpolation pixels currently compensated is also from one of interpolation pixels previously compensated, and the pixel bock compensated previously is located at the left side of the interpolation pixel, then the assigned value of the interpolation pixel is the value previous compensated; wherein, when the set directions are selected in a range from 90° to 180°, if the assigned value to one of the five interpolation pixels currently compensated is also from one of interpolation pixels previously compensated, and the pixel block compensated previously is located above the interpolation pixel, then the assigned value of the interpolation pixel is the value previously compensated, and if the assigned value to one of the five interpolation pixels currently compensated is also from one of interpolation pixels previously compensated, and the pixel block compensated previously is located at the left side of the interpolation pixel, then the assigned value of the interpolation pixel is the value currently compensated; and wherein, when the set directions are selected from outside of a range of 45° to 180°, if the assigned value to one of the five interpolation pixels currently compensated is from one of interpolation pixels previously compensated, and the pixel block compensated previously is located above or at the left side of the interpolation pixel, then the assigned value of the interpolation pixel is the value peviously compensated.

12. The system according to claim 11 , wherein, the edge detection means further detect the edges of each low-resolution image to be processed from four different directions by adopting a Sobel operator.